Ho Chi Minh City University of Technology and Education (Abbreviation: HCMUTE, Vietnamese: Trường Đại học Sư phạm Kỹ thuật Thành phố Hồ Chí Minh) is currently listed as one of the top 10 universities in Vietnam and also a member in the top group of Southeast Asia universities (basing on standard evaluation index).This is a public university located in Thủ Đức City, about 10 km (6.2 mi) north-east from downtown Ho Chi Minh City. This university offers bachelor's and associate degree to prospective lecturers in other technical institutions. The university also conducts technical research and vocational training, in addition to educational cooperation with foreign universities.
Interpretability is essential in Whole Slide Image (WSI) analysis for computational pathology, where understanding model predictions helps build trust in AI-assisted diagnostics. While Integrated Gradients (IG) and related attribution methods have shown promise, applying them directly to WSIs introduces challenges due to their high-resolution nature. These methods capture model decision patterns but may overlook class-discriminative signals that are crucial for distinguishing between tumor subtypes. In this work, we introduce Contrastive Integrated Gradients (CIG), a novel attribution method that enhances interpretability by computing contrastive gradients in logit space. First, CIG highlights class-discriminative regions by comparing feature importance relative to a reference class, offering sharper differentiation between tumor and non-tumor areas. Second, CIG satisfies the axioms of integrated attribution, ensuring consistency and theoretical soundness. Third, we propose two attribution quality metrics, MIL-AIC and MIL-SIC, which measure how predictive information and model confidence evolve with access to salient regions, particularly under weak supervision. We validate CIG across three datasets spanning distinct cancer types: CAMELYON16 (breast cancer metastasis in lymph nodes), TCGA-RCC (renal cell carcinoma), and TCGA-Lung (lung cancer). Experimental results demonstrate that CIG yields more informative attributions both quantitatively, using MIL-AIC and MIL-SIC, and qualitatively, through visualizations that align closely with ground truth tumor regions, underscoring its potential for interpretable and trustworthy WSI-based diagnostics
Video streaming in vehicular ad-hoc networks (VANETs) faces significant challenges due to the dynamic nature of vehicles, frequent disconnections, and huge demands for high data rate communications. These challenges make it longer for vehicle users (VUs) to complete their sessions in video applications and services (VASs). In this paper, we fully utilize the benefits of both deterministic and probabilistic edge caching (DPC) techniques for cooperative transmission to minimize the service time in VASs. To do so, a DPC optimization problem is formulated and solved for the optimal results of 1) caching placement in roadside units (RUs) and 2) caching probability in VUs under the constraint on caching storage resource. Genetic algorithms are modified to deal with the complexity of two types of optimization variables, i.e., integer variable for deterministic caching and real variable for probabilistic caching, and thus ensuring high stability and accuracy. Simulation results demonstrate that the DPC method outperforms the other conventional schemes in terms of service time while efficiently utilizing the storage of RUs and VUs. Important findings are also analyzed and discussed to provide more useful insights into the design of edge caching techniques for VASs in VANETs.
The growing demand for spectrum efficiency in next-generation wireless networks, especially in vehicular environments, necessitates effective spectrum sensing (SS) techniques capable of managing the coexistence of technologies like fifth generation new radio (NR) and radar systems. This letter introduces SpecDiff, an innovative framework based on latent diffusion models for spectrogram segmentation, designed to identify and differentiate these coexisting signals in dynamic, noisy environments. SpecDiff leverages a generative diffusion model in a compact latent space, using an attention-based denoising process to enhance segmentation performance under low signal-to-noise ratios and complex channel conditions. The model achieves state-of-the-art performance, with a mean accuracy of 98.68% and mean intersection-over-union (IoU) of 96.30%, effectively identifying the occupied bandwidth in spectrograms. Furthermore, SpecDiff surpasses existing deep learning models in both accuracy and efficiency, offering a promising solution for spectrum sharing in future wireless networks.
In the precision molding of transparent Polycarbonate (PC) electronic housings, manufacturers face an inherent conflict: the high processing temperatures required to minimize visible weld-line defects often trigger volumetric shrinkage that compromises the strict dimensional tolerances necessary for complex assembly. This study addresses this gap by presenting a novel, data-efficient optimization framework that simultaneously resolves these traditionally conflicting objectives under real industrial manufacturing conditions. Utilizing a Taguchi L16 orthogonal array integrated with multi-objective Grey Relational Analysis (GRA) and average Signal-to-Noise Ratio (SNR) consistency checks, the process reached an optimal configuration (70 degrees C mold temperature, 305 degrees C melt temperature, 75 mm/s fill velocity, 35 MPa holding pressure, and 20 s cooling time). Validation experiments achieved a significant reduction in weld-line width (WLW) of 57.00%, while dimensional deviations were improved by 85.92% and 50.90% compared to initial averages. Analysis of Variance (ANOVA) identified holding pressure (26.49%) and fill velocity (25.74%) as the dominant drivers for dimensional precision and overall aesthetic quality, respectively. This approach provides a robust, cost-effective guideline for precision manufacturing, ensuring that aesthetic enhancements do not hinder the mechanical fit in high-complexity electronic assemblies.
Residual particles generated during automotive transmission manufacturing can compromise system reliability, functional stability, and driving safety if they are not effectively removed before assembly. Those particles – burr, cast, chip, debris - can be generated from casting, CNC milling, CNC turning, high pressure water jet deburring, brushing deburring or unknown sources. Identifying the production origin of these particles is therefore essential for cleanability control and corrective action in manufacturing. This study proposes a morphology-based particle-classification framework using an artificial neural network (ANN) and ten extracted descriptors, including geometric features, fractal dimension, and curvature-related parameters. A two-stage classification strategy was developed to evaluate performance under different levels of classification difficulty. In Stage A, particles in the 300–1000 μm range were classified into seven representative classes, while five difficult subclasses were grouped into an “Unknown” category. Under this practical screening setting, the ANN achieved an overall accuracy of 91.8
